'Woosh' continuous noise removal on 33 LPs

[dcblana]

Hi there,

I have the recurring problem when restoring some LPs of a periodic 'woosh' sound in the hiss when using LIVE 6 FORENSICS. The continuous noise filter does not seem to do much to remove/reduce it unless I am not using the right settings/presets or the right filters. I have searched the help doc and could not find anything (Maybe I do not know the correct technical term for that type of noise)
Can anyone help me with this? Thanks.

Sincerely,
Dan Blana, Master's Touch Music

[Craig Maier]

I would recommend using the Auto CNF filter rather than the regular CNF (this setting is one of the "CNF Modes" that you can select). Also, I would keep the attenuation control as low as possible to reduce the noise but not introduce digital artifacts into the restoration.

The Auto CNF will be able to follow the noise inconsistencies on the recording much better than the other modes of the CNF filter because that mode is adaptive.

Another one worth trying is the Forensics AFDF (Adaptive Frequency Domain Filter). Again, like with the Auto CNF, adjust the attenuation for the lowest setting to achieve some noise reduction. Do not over do it.

[cromag]

Just out of curiosity, are there any obvious advantages/disadvantages to making one CNF pass using a "noise" sample and then a second pass with Auto CNF?

[Craig Maier]

That is an interesting question; I do not know the answer. I have never tried it.

[Dan McDonald]

I've done that with particularly nasty material and it can help. I think the order I used most commonly was CNF first to get rid of the basic noise, then the auto CNF. I have used it the other way occasionally also, with Auto first and the CNF to get rid of some artifacts. The second order helps when you have to be more aggressive than you want to be, so you introduce some artificats. Sometimes the spectral subtraction works for the second one in that situation.

The main situations I use multipe CNF in is when there are multiple copies of a tape, because there is noise on top of noise.

When people copy tapes, a lot of times they start the recorder first, then start the playback machine, so that they get it all. In that case, you get a layer of noise of the copying recorder first, which you can sample, and then a layer of noise of the original on top of the copying recorder. I like to get rid of copier's noise, then work the original noise.

I don't know if it's any better or not, but it works better for me that way.

I'd imagine that would work for lps that are collections of 78s, which would have lp noise and the original 78 noise.

Dan

[Craig Maier]

It is theoretically possible that using different values of FFT size for each of the two passes might help. But, that is in theory only. Again, I have never tried any of this. But, it is interesting, especially Dans observation about the layers of noise that can exist on taped sources.

[Craig Maier]

On further reflection on the use of dual CNF's, consider the tradeoff between FFT size as it pertains to transient response vs. frequency selectivity. Large FFT sizes provide superior frequency selectivity at the expense of transient response. Visa versa is also true, meaning that small FFT sizes provide improved transient response at the expense of poorer frequency selectivity.

So, perhaps an effective methodology could involve "dividing the noise in half" and attacking each half seperately. What I mean by that is to not try to get all the noise in one pass; just get half of it. Get the other half with a second pass. Make the first pass have a small FFT size, but go real easy on the attenuation control. Get the rest of the noise with a second pass having a large FFT size, again keeping the attenuation control at a small setting.

As a sidebar, this could all be set up in the multifilter to enjoy the benefits provided by the multifilter methodology.

Lastly, I still have not tried this, so it is all theory, which could turn out to be misleading.

Remember the olde addage:

One test is worth a million expert opinions!